Upload 3 files
Browse files
model_assets/Rinne/Rinne.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:72d8969b3883f5abd22ba7afd8d9a9353f04f78199f1e340d472c8bdd8ee3c59
|
3 |
+
size 251150980
|
model_assets/Rinne/config.json
CHANGED
@@ -5,29 +5,34 @@
|
|
5 |
"eval_interval": 2000,
|
6 |
"seed": 42,
|
7 |
"epochs": 1000,
|
8 |
-
"learning_rate": 0.
|
9 |
"betas": [
|
10 |
0.8,
|
11 |
0.99
|
12 |
],
|
13 |
"eps": 1e-09,
|
14 |
"batch_size": 3,
|
15 |
-
"bf16_run":
|
16 |
-
"
|
|
|
17 |
"segment_size": 16384,
|
18 |
"init_lr_ratio": 1,
|
19 |
"warmup_epochs": 0,
|
20 |
"c_mel": 45,
|
21 |
"c_kl": 1.0,
|
|
|
22 |
"skip_optimizer": false,
|
23 |
"freeze_ZH_bert": false,
|
24 |
"freeze_JP_bert": false,
|
25 |
"freeze_EN_bert": false,
|
26 |
-
"
|
|
|
|
|
27 |
},
|
28 |
"data": {
|
29 |
-
"
|
30 |
-
"
|
|
|
31 |
"max_wav_value": 32768.0,
|
32 |
"sampling_rate": 44100,
|
33 |
"filter_length": 2048,
|
@@ -39,19 +44,20 @@
|
|
39 |
"add_blank": true,
|
40 |
"n_speakers": 1,
|
41 |
"cleaned_text": true,
|
|
|
|
|
|
|
42 |
"num_styles": 1,
|
43 |
"style2id": {
|
44 |
"Neutral": 0
|
45 |
-
},
|
46 |
-
"spk2id": {
|
47 |
-
"Rinne": 0
|
48 |
}
|
49 |
},
|
50 |
"model": {
|
51 |
"use_spk_conditioned_encoder": true,
|
52 |
"use_noise_scaled_mas": true,
|
53 |
"use_mel_posterior_encoder": false,
|
54 |
-
"use_duration_discriminator":
|
|
|
55 |
"inter_channels": 192,
|
56 |
"hidden_channels": 192,
|
57 |
"filter_channels": 768,
|
@@ -99,7 +105,14 @@
|
|
99 |
],
|
100 |
"n_layers_q": 3,
|
101 |
"use_spectral_norm": false,
|
102 |
-
"gin_channels":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
103 |
},
|
104 |
-
"version": "1
|
105 |
}
|
|
|
5 |
"eval_interval": 2000,
|
6 |
"seed": 42,
|
7 |
"epochs": 1000,
|
8 |
+
"learning_rate": 0.0001,
|
9 |
"betas": [
|
10 |
0.8,
|
11 |
0.99
|
12 |
],
|
13 |
"eps": 1e-09,
|
14 |
"batch_size": 3,
|
15 |
+
"bf16_run": false,
|
16 |
+
"fp16_run": false,
|
17 |
+
"lr_decay": 0.99996,
|
18 |
"segment_size": 16384,
|
19 |
"init_lr_ratio": 1,
|
20 |
"warmup_epochs": 0,
|
21 |
"c_mel": 45,
|
22 |
"c_kl": 1.0,
|
23 |
+
"c_commit": 100,
|
24 |
"skip_optimizer": false,
|
25 |
"freeze_ZH_bert": false,
|
26 |
"freeze_JP_bert": false,
|
27 |
"freeze_EN_bert": false,
|
28 |
+
"freeze_emo": false,
|
29 |
+
"freeze_style": false,
|
30 |
+
"freeze_decoder": false
|
31 |
},
|
32 |
"data": {
|
33 |
+
"use_jp_extra": true,
|
34 |
+
"training_files": "Data\\Rinne\\train.list",
|
35 |
+
"validation_files": "Data\\Rinne\\val.list",
|
36 |
"max_wav_value": 32768.0,
|
37 |
"sampling_rate": 44100,
|
38 |
"filter_length": 2048,
|
|
|
44 |
"add_blank": true,
|
45 |
"n_speakers": 1,
|
46 |
"cleaned_text": true,
|
47 |
+
"spk2id": {
|
48 |
+
"Rinne": 0
|
49 |
+
},
|
50 |
"num_styles": 1,
|
51 |
"style2id": {
|
52 |
"Neutral": 0
|
|
|
|
|
|
|
53 |
}
|
54 |
},
|
55 |
"model": {
|
56 |
"use_spk_conditioned_encoder": true,
|
57 |
"use_noise_scaled_mas": true,
|
58 |
"use_mel_posterior_encoder": false,
|
59 |
+
"use_duration_discriminator": false,
|
60 |
+
"use_wavlm_discriminator": true,
|
61 |
"inter_channels": 192,
|
62 |
"hidden_channels": 192,
|
63 |
"filter_channels": 768,
|
|
|
105 |
],
|
106 |
"n_layers_q": 3,
|
107 |
"use_spectral_norm": false,
|
108 |
+
"gin_channels": 512,
|
109 |
+
"slm": {
|
110 |
+
"model": "./slm/wavlm-base-plus",
|
111 |
+
"sr": 16000,
|
112 |
+
"hidden": 768,
|
113 |
+
"nlayers": 13,
|
114 |
+
"initial_channel": 64
|
115 |
+
}
|
116 |
},
|
117 |
+
"version": "2.4.1-JP-Extra"
|
118 |
}
|